Gibbs Artifact Reduction Using Local Subpixel Shift And Interlaced Local Variation

Gibbs artifacts are a ring-shaped artifact commonly found at the edge of MRI that is common in Magnetic Resonance Imaging(MRI).In clinical applications,taking into account the actual situation of the patient,a certain number of k-space peripheral phase encoding lines are usually ignored to shorten the acquisition time.The reconstruction of the finite data using the Fourier transform results in Gibbs artifacts in the reconstructed image,which ultimately affects the diagnostic value of clinical MRI.How to eliminate the Gibbs artifacts present in clinical magnetic resonance images becomes very important.Local Subpixel Shift(LSS)proposed by Elias et al.eliminates Gibbs artifacts by searching for the smallest local total variation in the image domain.Although this method eliminates the artifacts,the method is not applicable to k-space zero-padding reconstructed data.In the first part of this paper,we present two methods for applying LSS to under sampled data.The first method is LSS+ interpolation.The method firstly transforms the non-zero-filled k-space data into the image domain,and then uses the LSS algorithm to search for the smallest sub-pixel shift in the image domain.Finally,the image is interpolated.The final image.The second method is the interlaced local variation method,which first zero-fills the k-space data,and then transforms the zero-padded k-space data into the image domain,and then interlaces the local variation in the image domain(interlaced).The Local Variation,iLV)method searches for the best LSS,which reduces the Gibbs artifacts of the image.In order to compare the effectiveness of the two methods presented in this paper,we compare the iLV and LSS+ interpolation with the original LSS algorithm and the Hamming window filter.The feasibility of the two methods is verified in the phantom and T2WI data.Sex and robustness.We find that the iLV method can effectively eliminate Gibbs artifacts in images compared with iLV and LSS proposed in this paper.Compared with LSS+ interpolation method,iLV method can better preserve the details of images;and Hamming window filtering method In contrast,the Hamming window filtering method can lead to great edge smoothing.In the second part of the paper,we use the iLV method in the g-space zero-fill reconstruction tensor imaging Gibbs artifact data to study the improvement of Gibbs artifacts in the diffusion tensor data by the iLV method.First,we preprocessed the collected diffusion tensor data of C57BL/6J mice,and then performed different degrees of zero-filling on the pre-processed data to simulate the diffusion tensor data containing Gibbs artifacts,and then use the iLV method.The diffusion tensor data containing Gibbs artifacts are corrected,the diffusion tensor calculation is performed on the data before and after the correction,the statistical analysis of the brain region of interest is performed on the parameter map,and the brain of interest is reconstructed by the deterministic fiber tracking algorithm.The fiber bundle of the area.We found that there was a significant difference in the FA values of the brain regions of interest in the positive and negative data before and after the iLV method.In addition,we also performed deterministic fiber tracking of the brain region of interest in the k-space 50%zero-filled reconstructed diffusion tensor data.The results show that the fiber bundles tracked by the iLV method will have a greater density of fiber bundles at the cingulate back.The fiber length is also greatly improved.The iLV method proposed in this paper implements the extension of the traditional LSS method,and can be applied to eliminate Gibbs artifacts in the k-space zero-fill reconstructed image,and can preserve the detailed information of the image well.